Stream Conversation AI Agent Response (SSE)

⚠️ This is a streaming endpoint using Server-Sent Events (SSE). Opens a live SSE stream to receive real-time response messages from the conversational agent after sending a message. Use the temporary token received from /generate_secret or /get_inference as applicable to authenticate the stream. The stream sends event.data messages that include control signals or agent-generated content.

1. Connection Closing

close_connection/[DONE]: Indicates the normal end of the stream. The client must close the EventSource connection.

2. Stream Continuation

call_again: Indicates a tool/function call occurred. You’ve to set your message loading state to true and new message chunks will be sent in the stream.

3. Error / Restriction Signals

These indicate problems or user restrictions. In all cases, the client must close the connection.

no_message_found: The agent was unable to generate a response as no user message corresponding to request found.
close_connection_incomplete_content: AI model exceeded its response context window.
close_connection_incomplete_tool_calls: Context overflow occurred while generating tool call arguments.
close_connection_incomplete_request: Context overflow occurred while generating response.
__ERROR__: ....: An error occurred while processing the request.
[FROZEN] / [BLACKLISTED]: User access restrictions triggered.

4. AI Agent Response Content

JSON payloads: Contain AI Agent Response message content. When streaming is ON you will receive message chunks, and when streaming is OFF you will get the full response in a single JSON payload.

When Streaming is ON:

id: 725e1f05-6142-467e-bdec-47862d50da79
data: {"content": "How"}

id: 725e1f05-6142-467e-bdec-47862d50da79
data: {"content": " can"}

id: 725e1f05-6142-467e-bdec-47862d50da79
data: {"content": " I"}

id: 725e1f05-6142-467e-bdec-47862d50da79
data: {"content": " assist"}

id: 725e1f05-6142-467e-bdec-47862d50da79
data: {"content": " you"}

id: 725e1f05-6142-467e-bdec-47862d50da79
data: {"content": " today"}

id: 725e1f05-6142-467e-bdec-47862d50da79
data: {"content": "?"}

id: 725e1f05-6142-467e-bdec-47862d50da79
data: {"signal": "close_connection", "new_message_time": "2025-09-04T08:55:06.298621", "new_message_id": "44cabb53-1253-46db-8c77-d36fa098773f"}

data: [DONE]

When Streaming is OFF:

id: 725e1f05-6142-467e-bdec-47862d50da79
data: {"content": "How can I assist you today?"}

data: {"signal": "close_connection", "new_message_time": "2025-09-04T08:55:06.298621", "new_message_id": "44cabb53-1253-46db-8c77-d36fa098773f"}

data: [DONE]

Example Implementation

const displayNewMessageHandler = (id, message) => {
  setConversationMessages((prev) => {
    // set your messages state here
    const lastMessage = prev[prev.length - 1];
    if (lastMessage.role === 'assistant' && lastMessage.id === id) {
      // same id -> chunks of message is to be appended
      // last message is assistant and has the same token -> chunks of message is to be appended
      const updatedData = [...prev];
      updatedData[prev.length - 1] = {
        ...lastMessage,
        content: lastMessage.content + message,
      };
      return updatedData;
    } else {
      // new assistant message
      const newMessage = { role: 'assistant', content: message, time: new Date().toLocaleString('en-IN'), id: id };
      return [...prev, newMessage];
    }
  });

  // you have to save the conversationMessages to local storage or any other persistent storage, chatzy AI doesn't expose any API to fetch conversation history
};

const callSourceHandler = (token) => {
  // token is the temp token retrieved by `/get_inference`
  const source = new EventSource(`https://backend-1.chatzy.ai/get-response?token=${token}`);

  source.onmessage = (event) => {
    const id = event.lastEventId;

    if (event.data === '[DONE]') {
      // End of stream
      source.close();
    } else if (event.data === 'no_message_found') {
      source.close();
    } else if (['[FROZEN]', '[BLACKLISTED]'].includes(event.data)) {
      source.close();
    } else if (event.data.includes('__ERROR__')) {
      const errorMessage = event.data.replace('__ERROR__:', '');
      console.error('Error occurred while processing the request:', errorMessage);
      source.close();
    } else {
      let dataObject;
      try {
        dataObject = JSON.parse(event.data);
      } catch {
        // console.error('Error parsing JSON data:', event.data);
      }

      if (dataObject?.content) {
        displayNewMessageHandler(id, dataObject.content);
      } else if (dataObject?.signal === 'close_connection') {
        source.close();
      } else if (dataObject?.signal === 'call_again') {
        // Tool call executed
        // show your message loading
      } else if (dataObject?.signal === 'close_connection_incomplete_content') {
        source.close();
        // show error message to user
      } else if (dataObject?.signal === 'close_connection_incomplete_tool_calls') {
        source.close();
        // show error message to user
      } else if (dataObject?.signal === 'close_connection_incomplete_request') {
        source.close();
        // show error message to user
      }
    }
  };

  source.onerror = function (error) {
    source.close();
    console.error(`EventSource failed:`, error);
  };
};

Query Parameters

token

string

required

Temporary access token used to authenticate the Server-Sent Events (SSE) stream.

This token is obtained from the /generate_secret endpoint and also after sending a message to the AI agent via /get_inference.

Pass it as a query parameter like this:

/get-response?token=YOUR_TEMP_TOKEN

If the token is missing or invalid, the server will respond with a 401 Unauthorized error.

Example:

"eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9..."

Response

200 - text/event-stream

SSE stream opened. Messages will be sent continuously as text/event-stream.

The response is of type string.

Example:

"data: {\"content\": \"How\"}\n\ndata: {\"content\": \" can\"}\n\ndata: {\"content\": \" I\"}\n\ndata: {\"content\": \" assist\"}\n\ndata: {\"content\": \" you\"}\n\ndata: {\"content\": \" today\"}\n\ndata: {\"content\": \"?\"}\n\ndata: close_connection\n\n"

Documentation Index

​1. Connection Closing

​2. Stream Continuation

​3. Error / Restriction Signals